CN103678101A - Method for detecting software reliability of high-speed train network control system - Google Patents

Abstract

The invention relates to a method for detecting the reliability of a high-speed train network control system. The method includes the steps that a maximum likelihood estimation method is used for obtaining model parameters through error occurrence time data obtained in the reliability testing process and according to a Jelinski-Moranda model formula; the model parameters are substituted into the model formula to obtain results of all reliability parameters of the system, so that whether the system meets the reliability requirements or not is judged. According to the method, software reliability parameters can be evaluated quantitatively through the error occurrence time data obtained in the testing process and by the utilization of a reliability growth model derivation formula, meanwhile how much testing time is still needed or how many questions needs to be found out for enabling software to meet the reliability requirements can be estimated, and reference is provided for testing personnel or managers for testing schedule control and project schedule grasp. A detection system of the method is unrelated to a used programming language. Detection of the software system reliability is not directly related to the programming language which is selected for authoring the software.

Description

The software reliability detection method of bullet train network control system

Technical field

The present invention relates to the reliability detection technique of software, particularly can detect for the reliability index of bullet train network control system quantitative the method for the reliability of bullet train network control system.

Background technology

Software is world today's industry with the fastest developing speed, yet has also brought thus a very difficult problem---the quality and reliability problem of software.The rapid expanding of software size, makes its difficult quality guarantee.The software crisis that 20 century 70s cause, has facilitated the birth of soft project.

Bullet train network control system is the intensive system of software, and the q&r of software is most important.Along with the develop rapidly of computer hardware technique, for meeting the needs of modern train technical development, bullet train uses computer system in a large number, and many computer systems play vital effect for the reliability, the security that guarantee train operation.And also day by day receive people's concern as the quality of the computer software of its core.The thrashing causing due to computer software defect may lead to disastrous consequence.

Three key properties of software product are quality, expense and progress.The object that proposes soft project is exactly under certain expense and progress restriction, to develop high-quality software product.Very easy to the tolerance of expense, progress, but but very difficult to the tolerance of software quality, but extremely important.

Software quality comprises six characteristics: functional, reliability, ease for use, efficiency, maintainability and portability.Wherein software reliability is most important characteristic, is also the characteristic of easy quantitative measurement.

The degree of reliability of software need measure to represent by software reliability, and by quantizating index, examine software reliability level, Common Parameters comprises: fiduciary level, crash rate and mean time between failures etc.Parameters is defined as follows:

(1) fiduciary level

Software reliability R refers to that software, under defined terms, completes the probability of predetermined function in official hour section.Or perhaps the software probability that no-failure occurs in the given time.

If official hour section is t ₀, the time that software occurs to lose efficacy is ξ:

Fiduciary level R (t ₀)=p (ξ > t ₀)

Unreliable degree F (t ₀)=1-R (t ₀)=P (ξ≤t ₀)

Failure dense function f (t)=dF (t)/dt

(2) crash rate and failure intensity

Crash rate refers in the situation that t not yet occurs to lose efficacy constantly, at t, the probability losing efficacy occurred in the unit interval constantly.

$Z\left(t\right)=\underset{\mathrm{\&Delta;t}\&RightArrow;0}{\lim }\frac{P(t<\mathrm{\&xi;}<t+\mathrm{\&Delta;t}|\mathrm{\&xi;}>t)}{\mathrm{\&Delta;t}}=\frac{f\left(t\right)}{R\left(t\right)}$

Failure intensity: suppose that the failure number that software occurs constantly at t is M(t) be, obvious M(t) random number, and the variation of t in time and difference, M(t), t > 0} is a stochastic process.If μ (t) is stochastic variable M(t) average, have μ (t)=E[M(t)], λ (t)=d μ (t)/dt is t failure intensity constantly.

Crash rate and failure intensity are two different concepts, but closely related.Crash rate (failure rate) is also translated into failure rate sometimes, and its definition is consistent with crash rate definition in hardware reliability, is that the viewpoint based on the life-span provides, and is a conditional probability density.And failure intensity is based on stochastic process definition, it is the rate of change of failure number average.In hardware reliability, crash rate represents with λ, and in software reliability, λ is used for representing failure intensity.In order to distinguish, crash rate represents with Z.In some document, also crash rate is called to level of significance (hazard rate).

In some situation, software failure rate and failure intensity are equal, can be mutually general.M(t), and when t > 0} is a Poisson process, conditional failure rate functions Z (Δ t|t _i) and failure intensity function lambda (Δ t+t _i) be identical; If under the stable condition of using software and software not being changed, the failure intensity of software should be a normal value, and: λ=Z=is often worth, at this moment M(t), t > 0} is a homogeneous Poisson processes.In the present invention, crash rate and failure intensity are not distinguished.

(3) mean time to failure MTTF/ mean time between failures MTBF

The current time that refers to MTTF arrives the average of out-of-service time next time.MTBF refers to the average at twice adjacent out-of-service time interval.In hardware reliability, MTTF is for non-repairable item, and MTBF is for repairable item; For software, can not by same concept, distinguish simply, because there is not the inefficacy of unrepairable in software.In software, the two difference is that the event object of research is TTF or TBF.TTF refers to from current time and brings into operation software to the time that occurs that lost efficacy next time, and TBF refers to the time interval of losing efficacy last time and losing efficacy to next time, this time interval has gone out to comprise restarting systems until outside the time of occur losing efficacy next time strictly speaking, also should comprise before restarting systems such as working times such as crash handlings.But this portion of time generally accounts for quite little ratio in whole program runtime, if do not consider the working time that this is a part of, TTF and TBF are on all four.In the present invention, do not distinguish two concepts, can be mutually general.

In prior art, system reliability parameter and computing method that bullet train network control system explicitly calls for have following:

1. crash rate λ (failure rate)

The relation of the number of times x occurring by fault and quantity n and the corresponding observation time of considered same unit, calculate the failure rate with time correlation:

${\mathrm{\&lambda;}}_s=\frac{x}{n*t}$

When a certain system is tested, can choose n this system and test simultaneously, record n test duration t and the fault frequency x that system is total, the above-mentioned formula of substitution.

The number of times x occurring by fault and the quantity n of considered same unit and the corresponding relation of observing mileage, calculate the failure rate relevant to mileage:

${\mathrm{\&lambda;}}_s=\frac{x}{n*s}$

2.MTBF/MDBF(mean failure rate mileage)

Suppose that a kind of failure condition is the constant to the time, utilize formula below to calculate MTBF:

$\mathrm{MTBF}=\frac{1}{{\mathrm{\&lambda;}}_s}$

In like manner try to achieve MDBF: $\mathrm{MDBF}=\frac{1}{{\mathrm{\&lambda;}}_s}$

By average velocity V _d, time correlation dependability parameter and mileage related reliability parameter can be changed mutually.

V _D＝s/t

Wherein S is average distance travelled, and t is average operating time.

MDBF＝MTBF×V _D

${\mathrm{\&lambda;}}_s=\frac{{\mathrm{\&lambda;}}_t}{V_D}$

The reliability index requiring is:

For 1 class fault, MDBF=400.000km; For 2 class faults, MDBF=100.000km.

Summary of the invention

If use the detection method of above-mentioned formula because needs complete reliability Complete test, expend a large amount of time and manpower and materials, in order to solve deficiency of the prior art, this method provides a kind of new method of carrying out reliability detection towards bullet train network control system, by the wrong time of origin data of obtaining in reliability testing process, according to the formula of Jelinski-Moranda model, use Maximum Likelihood Estimation to obtain model parameter, the result of each dependability parameter of substitution model formation acquisition system judges whether system meets reliability requirement again.

To achieve these goals, it is because this model hypothesis condition is consistent with the assumed condition of original requirement that the present invention adopts the reason of Jelinski-Moranda model, is all that assumed fault is the constant to the time.The method can replace original formula detection method under certain conditions, saves a large amount of time and manpower and materials, raises the efficiency.About the detailed introduction of Jelinski-Moranda model, can be with reference to < < Handbook of Software Reliability Engineering > > chapter 3 or other pertinent literature; About the detailed introduction of Maximum Likelihood Estimation, < < Probability Theory and Math Statistics > > the 7th chapter or other teaching material about probability statistics that can publish with reference to publishing house of Tsing-Hua University.

The present invention is towards the software reliability detection method of bullet train network control system, and its step comprises:

1) read the fail data in train network control system, described fail data comprises: current which inefficacy, inefficacy relative time and User Defined inefficacy severity level, and be saved to this locality with TXT textual form;

2) to described fail data at the lower Jelinski-Moranda model of setting up dependability parameter that imposes a condition;

3) according to single crash rate density function on any time in described model and the unreliable degree of single inefficacy, obtain failure intensity function and inefficacy mean value function in this network control system;

4) according to Maximum Likelihood Estimation Method, the parameter in described function is estimated;

5) estimated value is returned to computing formula in described model, obtained reliability judgement parameter, detect current results and whether meet the reliability conditions of setting.

The described Jelinski-Moranda of foundation model impose a condition for:

When 2-1) identical for the environment for use of unknown constant mistake sum N and test environment in software and expection, this model-composing condition meets: the possibility seriousness rank crash rate identical and separate and/or program identical, User Defined inefficacy that all inefficacies occur is constant in each time between failures;

If 2-2) meet step 2-1) in impose a condition, described model-composing meets: when mistake being detected in test and being repaired, the misarrangement time ignores and do not introduce new wrong and/or described crash rate numerical value and is proportional to residual errors number in program.

In described Jelinski-Moranda model, in each out-of-service time, crash rate is: Z (t)=Ф (N-i+1), and wherein Ф is the crash rate of single inefficacy, t is the i-1 time time interval of losing efficacy and losing efficacy to the i time.

According to crash rate Z (t) in described each out-of-service time, obtain: the unreliable degree F of single inefficacy _a(t)=1-exp (Ф t), single crash rate density function f _a(t)=Ф exp (Ф t), single crash rate Z _a(t)=Ф.

In described Maximum Likelihood Estimation Method based on time between failures x=(x ₁, x ₂... x _n) given subsample, set up the likelihood function of total wrong number N and single crash rate Ф.

Described estimated value is returned to reliability judgement parameter in described model calculation formula and is comprised:

Fiduciary level R (t)=exp[-Ф (N-i+1) t]; Crash rate Z (t)=Ф (N-i+1) and mean time between failures

wherein, t lost efficacy for the i-1 time to arbitrary moment between the i time inefficacy, and N is wrong sum, and i is the arbitrary integer between 0 to N, represents which fail data of current discovery.

Further, crash rate Z (t) in described model calculation formula and R (t) are made as to known quantity, obtain detecting the test duration t of software reliability needs and the fail data i of discovery.

Further, the parameter whether described detection current results meets reliability conditions is: a class fault, MDBF=400.000km; Two class faults, MDBF=100.000km.

Further, the chronomere of described fail data get minute, second, hour;

Further, in described fail data, fault type is divided into software fault and hardware fault.

Technique effect of the present invention is as follows:

The present invention is by the wrong time of origin data of obtaining in test process, the formula that dependability model of growth is derived, parameter that can qualitative assessment software reliability.

Can estimate also to need how many test durations or how many problems of finding just can make software reach reliability requirement, for tester or managerial personnel control testing progress, hold project process reference is provided.

Can graphically show the time dependent curve of software reliability.

The programming language of detection system of the present invention and use is irrelevant.The detection of software systems reliability and select what program language to write between software and do not contact directly.Meanwhile, also irrelevant with the software development methodology of specifically using.

Accompanying drawing explanation

Fig. 1 is the software reliability detection method process flow diagram of bullet train network control system of the present invention.

Fig. 2 is an embodiment operation steps process flow diagram in the present invention.

Fig. 3 is mistake-time curve graph of a relation of software reliability variation tendency in bright software reliability detection method in one embodiment of the invention.

Embodiment

Inventive principle

The software reliability detection method process flow diagram of bullet train network control system, comprises the steps: as shown in Figure 1

Read fail data text, fail data is imported in system, for model analysis provides data.In this method, fail data file is kept in hard disk with txt textual form, and this method provides Reading text program to carry out file reading.Fail data content comprises that losing efficacy is current which inefficacy, the severity level of inefficacy relative time and inefficacy.Wherein front two data contents are necessity input data that define in software reliability engineering, the 3rd data content for oneself defining according to demand in the present invention, and inefficacy rank define method is as follows:

1 class fault: caused by serious technical matters, train must stop before arriving point of destination, or can not dispatch a car according to plan for the same reason, needs standby train or trailer.As can not be drawn, pressurized air supply complete failure or air damping complete failure etc.;

2 class faults: train can not continue operation and too early can not normally the runing of class car by destination time table again, need to consider the limited operation of motor train unit, but not need to drag vehicle.As hauling ability drops to 50%, the Anti-spin device fault of fire alarm system complete failure, a bogie etc.;

3 class faults: vehicle is runed under the limited limit of minimum, but the highest overall trip speed is at least higher than 100km/h.As heat supply, heating ventilation and air-conditioning partial fault;

4 class faults: vehicle is runed not reducing under the condition of the highest overall trip speed.Such fault can be revised after predetermined operation finishes, and predetermined operation is not had to substantial effect.

According to Jelinski-Moranda model, set up the computing formula of the dependability parameters such as crash rate, fiduciary level, mean time between failures.

This model hypothesis condition is as follows:

1) mistake that observation starts in rear software adds up to N, is a unknown constant;

2) possibility that all inefficacies occur is identical, separate, and the seriousness rank losing efficacy is identical;

3) mistake detecting in test is repaired at once, and the misarrangement time ignores, and does not introduce new mistake;

4) crash rate of program is constant in each time between failures, and its numerical value is proportional to wrong number residual in program, between i test section in, crash rate is: Z (X _i)=Ф (N-i+1), wherein Ф is a proportionality constant, is the crash rate of single inefficacy, X _ibeing the time interval of losing efficacy and losing efficacy to the i time for the i-1 time, is a stochastic variable;

5) test environment of software is identical with the environment for use of expection.

In the present invention owing to can not confirming whether software meets as above condition, so first suppose that software meets above all conditions, then the calculating of deriving, the legitimate reading comparison that the model calculation and test are obtained, if comparative result approaches, can think that software meets assumed condition substantially, can use the testing result of this method.

If t is arbitrary moment of losing efficacy for the i-1 time between the i time inefficacy, by crash rate:

$Z\left(t\right)=\frac{f\left(t\right)}{R\left(t\right)}=\frac{F^{\&prime;}\left(t\right)}{R\left(t\right)}=\frac{-R^{\&prime;}\left(t\right)}{R\left(t\right)}=\mathrm{\&Phi;}(N-i+1)$

Wherein, f (t) is failure dense function, and R (t) is fiduciary level.

${\&Integral;}_0^{t}Z\left(x\right)\mathrm{dx}={\&Integral;}_0^t\frac{-R^{\&prime;}\left(x\right)}{R\left(x\right)}\mathrm{dx}={\&Integral;}_0^t-d\left[\ln R\left(x\right)\right]=-\ln R\left(t\right)$

Wherein, t is arbitrary moment of losing efficacy for the i-1 time between losing efficacy to the i time,

By to z(x) integration derives itself and fiduciary level R(t) relational expression, thereby obtain the formula of fiduciary level below.The mistake sum of N for mentioning in hypothesis article one, i is the arbitrary integer between 0 to N, represents which inefficacy of current discovery.

Can obtain

$R\left(t\right)=\exp [-{\&Integral;}_0^{t}Z\left(x\right)\mathrm{dx}]=\exp [-\mathrm{\&Phi;}(N-i+1)t]$

F(t)＝1-exp[-Ф(N-i-1)t]

f(t)＝Ф(N-i+1)exp [-Ф(N-i-1)t]

$\mathrm{MTTF}={\&Integral;}_0^{\&Proportional;}R\left(t\right)\mathrm{dt}=-\frac{1}{\mathrm{\&Phi;}(N-i+1)}\exp [-\mathrm{\&Phi;}(N-i+1)t]{|}_0^{\&Proportional;}=\frac{1}{\mathrm{\&Phi;}(N-i+1)}$

While being constant by the known crash rate of deriving above, out-of-service time obeys index distribution, MTTB(or MTBF) with crash rate be reciprocal relation, this kind of hypothesis supposes it is consistent with index in project.

To this model, the unreliable degree F of single inefficacy _a(t)=1-exp (Ф t), crash rate density function f _a(t)=Ф exp (Ф t), crash rate Z _a(t)=Ф is identical to all inefficacies.By independent crash rate density, can obtain cumulative failure intensity:

λ(t)＝Nf _a(t)＝NФexp(-Фt)

In like manner can obtain the mean value function of cumulative failure number:

μ(t)＝NF _a(t)＝N[1-exp(-Фt)]

From above formula, if we can know the value of N and Ф, just can obtain the dependability parameter value that we are concerned about.And we cannot obtain the actual value of total wrong number N and proportionality constant Ф in reality, therefore need to select certain method to obtain their estimated value.In the present invention, select Maximum Likelihood Estimation Method to obtain estimated value.

The step of Maximum Likelihood Estimation Method is as follows:

1) write out likelihood function.If density function is f (x; θ), and x=(x ₁, x ₂... x _n) be a given subsample, likelihood function is:

$L(\mathrm{\&theta;};x)=\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}f(x_i;\mathrm{\&theta;})$

2) likelihood function is taken the logarithm after differentiate, obtain likelihood equation:

$\frac{\&PartialD;\ln L(\mathrm{\&theta;};x)}{\&PartialD;\mathrm{\&theta;}}=0$

If testing in this model and obtaining time between failures is x ₁, X ₂... x _n, the likelihood function of N and Ф is:

$L(N,\mathrm{\&Phi;})=\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&Phi;}(N-i+1)\exp [-\mathrm{\&Phi;}(N-i+1)x_i]$

Solve

$\frac{\&PartialD;\ln L(N,\mathrm{\&Phi;})}{\&PartialD;N}=0$ $\frac{\&PartialD;\ln L(N,\mathrm{\&Phi;})}{\&PartialD;\mathrm{\&Phi;}}=0$

$\widehat{\mathrm{\&Phi;}}=\frac{n}{\hat{N}\left({\mathrm{\&Sigma;}}_{i=1}^n{x}_i\right)-{\mathrm{\&Sigma;}}_{i=1}^n(i-1)x_i}$

$\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{1}{\hat{N}-(i-1)}=\frac{n}{\hat{N}(1/{\mathrm{\&Sigma;}}_{i=1}^n{x}_i)\left({\mathrm{\&Sigma;}}_{i=1}^n(i-1)x_i\right)}$

When

time likelihood equation without reasonable solution,

time have unique reasonable solution.

By the estimated value substitution Z (t) obtaining, R (t), the isoparametric computing formula of MTTF, can try to achieve the estimated value of these parameters.

If current result of calculation does not meet the requirements of index, (being 400.000 as mentioned a class fault requirement MDBF above) current results can not meet reliability requirement, the present invention can be to also needing how many test durations or how many problems of finding just can reach requirement reliability level etc. and predict, and graphically show the variation tendency of reliability with the test duration.In formula above using crash rate, reliability index as known quantity substitution formula, the conversely i in computing formula and t.

In order to make object of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, the present invention is further elaborated.Specific embodiment described herein only, in order to explain the present invention, is not intended to limit the present invention, and example used can be used the program of the exploitations such as C#, JAVA, C++ to simulate.

To illustrate for example native system technical scheme below the software reliability detection method operation steps process flow diagram of bullet train network control system of the present invention as shown in Figure 2.Following table is depicted as certain system test result, the information of 8 faults that record test obtains altogether.First row represents time between failures, the time occurring apart from a upper fault when this fault occurs, chronomere can according to actual conditions get minute, second, hour etc.; Secondary series represents fault type, and " 0 " represents that fault is software fault, and " 1 " represents that fault is hardware fault, and the reliability of carrying out specially software systems due to native system detects, and therefore without special circumstances, secondary series data should be all " 0 "; The severity level of fault is shown in the 3rd list, and concrete classification delimited according to demand.

Table 1

Row 1	Row 2	Row 3
			1.00000	0	1
1.50000	0	2
			2.00000	0	3
2.50000	0	1
			2.50000	0	1
3.00000	0	1
			4.00000	0	2
6.00000	0	1

Using upper table 1 test data as input, the method of mentioning according to the present invention, first by Maximum Likelihood Estimation Method, estimate proportionality constant Ф and total failure number N, as shown in following table 2 model analysis results, the value 0.0898 of Ф, the value of N is 9, then in the formula that the value substitution Jelinski-Moranda model of Ф and N is derived, can obtain each reliability index parameter value.Concrete outcome can see the following form 2.

Table 2

Dependability parameter	Result of calculation
		Proportionality constant	0.08098
Primary failure strength function	0.72986
		Current failure intensity function	0.11801
The current mean time between failures	12.19185
		Current mean distance between failures	2438.36980
Total failure number	9
		Residue failure number	1

Given predicted target values, can obtain index of correlation data according to model formation equally, and concrete outcome sees the following form and 3 predicts the outcome.

Table 3

As shown in Figure 3, with the curve of No. *, be the curve of authentic testing data fitting, the curve that is not model prediction with the curve of No. *.Curve horizontal ordinate is detected inefficacy number, and ordinate is total test duration.Mistiming between i inefficacy and i+1 inefficacy can be thought the current mean down time of system after i inefficacy being detected.If two curve is more approaching, illustrate that the effect of modeling is better, can think that system under test (SUT) meets model hypothesis, the testing result that can use a model.

Claims (10)

1. a software reliability detection method for bullet train network control system, its step comprises:

1) read the fail data in train network control system, described fail data comprises: current which inefficacy, inefficacy relative time and User Defined inefficacy severity level, and be saved to this locality with TXT textual form;

2) to described fail data at the lower Jelinski-Moranda model of setting up dependability parameter that imposes a condition;

3) according to single crash rate density function on any time in described model and the unreliable degree of single inefficacy, obtain failure intensity function and inefficacy mean value function in this network control system;

4) according to Maximum Likelihood Estimation Method, the parameter in described function is estimated;

5) estimated value is returned to computing formula in described model, obtained reliability judgement parameter, detect current results and whether meet the reliability conditions of setting.

2. the software reliability detection method of bullet train network control system as claimed in claim 1, is characterized in that, the described Jelinski-Moranda of foundation model impose a condition for:

When 2-1) identical for the environment for use of unknown constant mistake sum N and test environment in software and expection, this model-composing condition meets: the possibility seriousness rank crash rate identical and separate and/or program identical, User Defined inefficacy that all inefficacies occur is constant in each time between failures;

If 2-2) meet step 2-1) in impose a condition, described model-composing meets: when mistake being detected in test and being repaired, the misarrangement time ignores and do not introduce new wrong and/or described crash rate numerical value and is proportional to residual errors number in program.

3. the software reliability detection method of bullet train network control system as claimed in claim 1, it is characterized in that, in described Jelinski-Moranda model, in each out-of-service time, crash rate is: Z (t)=Ф (N-i+1), wherein Ф is the crash rate of single inefficacy, and t is the i-1 time time interval of losing efficacy and losing efficacy to the i time.

4. the software reliability detection method of bullet train network control system as claimed in claim 3, is characterized in that, according to crash rate Z (t) in described each out-of-service time, obtains: the unreliable degree F of single inefficacy _a(t)=1-exp (Ф t), single crash rate density function f _a(t)=Ф exp (Ф t), single crash rate Z _a(t)=Ф.

5. the software reliability detection method of bullet train network control system as claimed in claim 1, is characterized in that, in described Maximum Likelihood Estimation Method based on time between failures x=(x ₁, x ₂... x _n) given subsample, set up the likelihood function of total wrong number N and single crash rate Ф.

6. the software reliability detection method of bullet train network control system as claimed in claim 1, is characterized in that, described estimated value is returned to reliability judgement parameter in described model calculation formula and comprised:

Fiduciary level R (t)=exp[-Ф (N-i+1) t; Crash rate Z (t)=Ф (N-i+1) and mean time between failures

wherein, t lost efficacy for the i-1 time to arbitrary moment between the i time inefficacy, and N is wrong sum, and i is the arbitrary integer between 0 to N, represents which fail data of current discovery.

7. the software reliability detection method of bullet train network control system as claimed in claim 6, it is characterized in that, crash rate Z (t) in described model calculation formula and R (t) are made as to known quantity, obtain detecting the test duration t of software reliability needs and the fail data i of discovery.

8. the software reliability detection method of bullet train network control system as claimed in claim 1, is characterized in that, the parameter whether described detection current results meets reliability conditions is: a class fault, MDBF=400.000km; Two class faults, MDBF=100.000km.

9. the software reliability detection method of bullet train network control system as claimed in claim 1, is characterized in that, the chronomere of described fail data get minute, second, hour.

10. the software reliability detection method of bullet train network control system as claimed in claim 1, is characterized in that, in described fail data, fault type is divided into software fault and hardware fault.

Images